Selection of representative hyperspectral data and image pretreatment for model development in heterogeneous samples: A case study in sliced dry-cured ham

Sliced dry-cured ham arranged in ready-to-eat packages is a convenient and widely consumed commodity characterised by heterogeneity composition not only among different industrial batches but also through their horizontal vertical profiles, making precise nutrition labelling of the difficult task. Hyperspectral imaging techniques can serve as steadfast solution to predict overall major constituents visualise distributions. The main aim this study was define optimal protocol for pretreating hyperspectral images selecting representative data developing accurate predictive models excessively heterogeneous samples, using sliced case study. (400–1000 nm) were acquired homogeneous unsliced muscles. Partial least squares (PLS) regression fat, water, salt protein contents developed tested an independent dataset. PLS from whole surface most ones predicting with determination coefficient prediction (Rp2) 0.89, 0.85, 83 0.63 standard error (SEP) 1.43, 1.21, 0.51 1.57%, respectively. chemical resulting gave advantages technique over traditional methods spatial distribution within packaged slices.